Yes, the main thing is to decide on the best and least diruptive way forward. We're happy to wait for 2.0 for this to be fixed, but it certainly seems necessary, especially as we move away from exclusively indexing MARC/Library data.
We made the 110/111 locally for our last release without any side-effects. Here for example, you can see corporate authors in the facets: http://catalogue.nli.ie/Search/Results?lookfor=ireland&type=AllFields&submit=FIND
A particularly nasty example of the 700 problem was actually reported to us directly by an author: excluding the first result where he is the "Main Author", when searching for "Paul Gorry" as an Author, the book "Tracing Irish Ancestors" is ranked as the 6th book even though "Paul Gorry" is an equal co-author of the book: http://catalogue.nli.ie/Search/Results?lookfor=paul+gorry&type=Author. This also leads to labelling problems, where co-authors are unfairly relegated to "Other Authors" (in trunk) or "Contributors" (in our catalogue).
We'll open a Jira Ticket & perhaps look at this when moving to 2.0 Do you have any objections to making "author" multi-valued in 2.0?
I think that these are legitimate problems that are worth addressing. The limitation of “single main author,” in particular, has come to my attention as I try to make VuFind’s code more uniform – limitations of the MARC-inspired Solr schema are hard to reconcile with other systems like Summon which make different assumptions.
That being said, I’m not eager to dramatically change the Solr schema at this moment in time – since we’re on the cusp of VuFind 2.0, I’m trying to avoid disruptive changes to the code base until after the existing logic has been ported to the new architecture. I’d rather redesign this once rather than twice. We can certainly start discussing this, and it’s probably worth opening a JIRA ticket to collect feedback, but I would prefer to hold off a little longer on implementation. Of course, the fact that I’m reluctant to do the work right now doesn’t mean that I’m going to stop anyone else who feels like contributing a patch!
The initial option is a significantly less disruptive change and might be a good place to start work. I wonder if we could achieve this through a custom analysis chain and copyField in Solr in order to avoid having to change anything in the import process….
If you’re going to be on next week’s developers call, feel free to bring this up if you want to discuss some of these issues in real-time.
This SF email is sponsosred by:
Try Windows Azure free for 90 days Click Here
Vufind-tech mailing list